HPCwire: New InfiniBand Architecture to Incorporate Direct GPU Communication

Since 1986 - Covering the Fastest Computers
in the World and the People Who Run Them

Translation Disclaimer

Subscribe | Sign In

Visit additional Tabor Communication Publications

Leading HPC
Solution Providers

June 18, 2012

New InfiniBand Architecture to Incorporate Direct GPU Communication

HAMBURG, Germany, June 18 -- Mellanox Technologies, Ltd., a leading supplier of high-performance, end-to-end interconnect solutions for data center servers and storage systems, today announced support for the new generation of NVIDIA GPUDirect technology. NVIDIA GPUDirect technology dramatically accelerates communications between GPUs by providing a direct peer-to-peer communication data path between Mellanox’s scalable HPC adapters and NVIDIA GPUs, without transferring data to the CPU or server memory subsystem.

Without GPUDirect, GPU data must first be copied into the system memory before going over the network. Now, the interconnect and GPU are tightly linked with data exchanges directly between the two devices over a PCI Express 3.0 bus, completely bypassing the CPU and the system memory. This direct-connected approach enabled by GPUDirect delivers faster communications between GPUs across different systems by as much as 80 percent, while reducing the end-to-end latency.

“The high performance and compute density of GPUs have made them a compelling solution for computationally intensive HPC applications,” said Gilad Shainer, vice president of market development at Mellanox Technologies. “To ensure the highest level of application performance, scalability and efficiency, the communication between GPUs within a cluster must be performed as quickly as possible. GPUDirect enables NVIDIA GPUs and Mellanox ConnectX-3 adapters to provide an optimum GPU clustering technology.”

“Mellanox’s support for GPUDirect helps users maximize their cluster performance,” said Sumit Gupta, senior director of the Tesla business at NVIDIA. “The ability to transfer data directly to and from GPU memory dramatically speeds up system and application performance, enabling users to run computationally intensive code and get answers faster than ever before.”

GPU-based clusters are widely used for computationally-intensive tasks, such as seismic processing, computation fluid dynamics and molecular dynamics. Since the GPUs perform high-performance floating point operations over a very large number of cores, a high-speed interconnect is required to connect between the platforms to deliver the necessary bandwidth and latency for the clustered GPUs to operate efficiently and alleviate any bottlenecks in the GPU-to-GPU communication path.

Mellanox ConnectX based adapters are the world’s only InfiniBand solutions that provide full offloading capabilities critical to avoiding CPU interrupts, data copies and systems noise, while maintaining high efficiencies for GPU-based clusters. Combined with the availability of NVIDIA GPUDirect, Mellanox InfiniBand solutions are driving HPC environments to new levels of performance and scalability.

Supporting Resources:

Learn more about Mellanox FDR 56Gb/s InfiniBand adapters
Learn more about GPUDirect with Mellanox InfiniBand: Mellanox GPUDirect
Follow Mellanox on Twitter and Facebook

About Mellanox

Mellanox Technologies is a leading supplier of end-to-end InfiniBand and Ethernet interconnect solutions and services for servers and storage. Mellanox interconnect solutions increase data center efficiency by providing the highest throughput and lowest latency, delivering data faster to applications and unlocking system performance capability. Mellanox offers a choice of fast interconnect products: adapters, switches, software and silicon that accelerate application runtime and maximize business results for a wide range of markets including high performance computing, enterprise data centers, Web 2.0, cloud, storage and financial services. More information is available at www.mellanox.com.

-----

Source: Mellanox

Share Options

Subscribe

Subscribe to HPCwire

Discussion

There are 0 discussion items posted.

Join the Discussion

Become a Registered User Today!

Registered Users Log in join the Discussion

RSS Feeds

Feeds by Topic

Feeds by Industry

Feeds by Content Type

Subscribe to All Content

All

June 21, 2012

June 20, 2012

June 19, 2012

June 18, 2012

June 15, 2012

June 14, 2012

June 13, 2012

June 12, 2012

Feature Articles

Exascale Computing: The View from Argonne

As a result of the dissolution of DARPA's UHPC program, the driving force behind exascale research in the US now resides with the Department of Energy, which has embarked upon a program to help develop this technology. To get a lab-centric view of the path to exascale, HPCwire asked a three of the top directors at Argonne National Laboratory -- Rick Stevens, Michael Papka, and Marc Snir -- to provide some context for the challenges and benefits of developing these extreme scale systems.
Read more...

TOP500 Gets Dressed Up with New Blue Genes

The 39th TOP500 list was released today at the International Supercomputing Conference in Hamburg, Germany, with a new machine at the top. Sequoia, an IBM Blue Gene/Q machine, delivered a world record 16 petaflops on Linpack, knocking RIKEN's 10-petaflop K Computer into second place. The Japanese K machine had held the TOP500 title for a year.
Read more...

Intel Will Ship Knights Corner Chip in 2012

On Monday at the International Supercomputing Conference in Hamburg, Intel announced that Knights Corner, the company's first manycore product, would be in production before the end of 2012. The company also released a few more details about the upcoming product line, including the creation of a new Xeon brand for the architecture, some performance updates on pre-production silicon, and Cray's adoption of MIC as part of its future Cascade supercomputer.
Read more...